1 Group Id

2 Project Title
Defending Suspected Users by Exploiting Specific Distance Metric in Collaborative
Filtering Recommender Systems
3 Project Option
4 Internal Guide
5 Sponsorship and External Guide
6 Technical Keywords (As per ACM Keywords)
Recommender System, Collabrative Filtering, Schilling Attack, Nuke And Push Attack

7. Problem Statement
Existing methods are not able to differentiate between genuine and fake User. This paper
proposes method to improve the accuracy to differentiate attack and genuine user. The approach
consists of two stages: the stage of filtering out genuine users by exploiting suspected target
items and the stage of discriminating attackers by employing a new similarity metric. At the first
stage, we filter out a part of genuine users in order to reduce the computation time.
8. Abstract:A recommender system produces a list of suggestions for users based on their
preferences. It is an intelligent system that can help users to come across their interesting items.
Mostly In E-commerce websites, recommender systems employ the collaborative filtering
algorithm which is one of the most successful techniques. But, CFRSs are highly vulnerable to
shilling attacks or profile injection attacks. Previous methods lack the accuracy in
differentiating attack users from genuine users and have lot of computation time. This paper

proposes method to improve accuracy of differentiating attack and genuine user and improves
computation time too.
9. Goals and Objectives:

We develop a scheme for improving accuracy in discriminating attack user form genuine

user by proposing the detection algorithm.

We develop a scheme for outsourced data that takes into account both the security and
performance.

10. Relevant mathematics associated with the Project

Input: Data matrix M;

Empirical thresholds

Output: The detected result DR

Algorithm : Detection algorithm

To determine the target items, we design an absolute count threshold which pushes or
nukes the same items with the highest or lowest at least times. If the count for an item is
greater than , then the item susi is regarded as suspected target item. It is noteworthy that
there will be more false positives or false negative if is too small or large.

Mathematical formulation:
Detection rate is defined as the number of detected attack profiles divided by the number
of attack profiles.
detection rate = No of Detection / No of Attack Profiles
False alarm rate is the number of genuine profiles that are predicted as attack profiles
divided by the number of genuine profiles.
False alarm rate =No of False alarm / No of Genuine Profiles

Success Conditions: Fake profile injection can be stopped and we can provide unbiased
recommendation to user.

Failure Conditions: Due to schilling attack recommendation lists can be altered by fake
users.

11.Names of Conferences / Journals where papers can be published

12. Review of Conference/Journal Papers supporting Project idea

[1] C. Chung, P. Hsu, and S. Huang, A novel approach to filter out malicious rating profiles
from recommender systems Journal of Decision Support Systems, pp. 314325, 2013.
[2] F. Zhang and Q. Zhou, HHT-SVM: An online method for detecting profile injection attacks
in collaborative recommender systems, Knowledge-Based Systems, 2014.
[3] W. Zhou, Y. S. Koh, J. H. Wen, S. Burki, and G. Dobbie, Detection of abnormal profiles on
group attacks in recommender systems, Proceedings of the 37th international ACM SIGIR
conference on Research & development in information retrieval, pp. 955958, 2014.
[4] M. Morid and M. Shajari, Defending recommender systems by influence analysis,
Information Retrieval, pp. 137152, 2014.
[5] Z. Zhang and S. R. Kulkarni, Detection of shilling attacks in recommender systems via
spectral clustering, International Conference on Information Fusion, pp. 18, 2014.
[6] I. Gunes, C. Kaleli, A. Bilge, and H. Polat, Shilling attacks against recommender systems: A
comprehensive survey, Artificial Intelligence Review, pp. 133, 2012.
[7] Z. Zhang and S. Kulkarni, Graph-based detection of shilling attacks in recommender
systems, IEEE International Workshop on Machine Learning for Signal Processing, pp. 16,
2013.
[8] Z. A. Wu, Y. Q. Wang, and J. Cao, A survey on shilling attack models and detection
techniques for recommender systems, Science China, vol. 59, no. 7, pp. 551560, 2014.

[9] C. E. Seminario and D. C. Wilson., Attacking item-based recommender systems with power
items, ACM Conference on Recommender Systems, pp. 5764, 2014.
[10] J. Hu, D. C. Zhan, X. Wu, Y. Jiang, and Z. H. Zhou, Pairwised specific distance learning
from physical linkages, ACM Trans. Knowl. Discov. Data, 2014.
13. Plan of Project Execution